Anticoagulant plasma was collected from 83 lung cancer patients (59 untreated and 24 treated with neoadjuvant immuno-chemotherapy) from 2018 to 2021 at Beijing Chest Hospital and 91 healthy donors (controls) and stored at -80°C for analysis of blood sCD137 levels. In addition, immune cell subsets in blood samples from 29 untreated patients, 34 healthy donors and 10 fresh tumor tissues were analyzed.

Another retrospective cohort containing 90 NSCLC tissues collected in 2013 at Beijing Chest Hospital and their representative tissue microarrays (TMAs) were included. The TMAs were prepared using two 1-mm diameter tumor cores from the formalin-fixed paraffin-embedded (FFPE) tissue block of each patient. All involved patients had received no treatment before surgery and had histologically confirmed lung cancer. The patients were followed up until January 2020 to evaluate overall survival (OS).

Patient peripheral blood mononuclear cells (PBMCs) were isolated and resuspended at a density of 1×10⁶/ml in 10% RPMI 1640 medium (Gibco) containing 50 μg/ml phytohemagglutinin (PHA, Sigma) for 24 h, 48 h or 72 h. Total RNA was isolated from stimulated cells using an RNA Easy Fast Tissue/Cell Kit (Tiangen Biotech). Then, CD137 was amplified using the primers with the following sequences: sense, 5’-gactgttgctttgggacattta-3’; antisense, 5’-tcacatcctccttcttcttcttc-3’. The obtained PCR products were cloned into the T-easy vector (Promega Corporation) for sequencing.

To detect sCD137 mRNA in different cell subsets, Tregs, non-Tregs, CD4+ T cells, CD8+ T cells, and CD4+CD25- T cells were isolated by using a FACSAia III Cell Sorter (BD Bioscience). The concentration of all cell subset samples was made equal for ARMS-QPCR, which was performed with GoTaq^® Probe qPCR Master Mix (Promega Corporation) and Cobas Z480 (Roche) according to the manufacturer’s instructions. The sequences of the primers and probes were as follows: sCD137: sense, 5’-atctgtcgaccctggacaaaga-3’; antisense, 5’-tcacatcctccttcttcttctt-3; probe, 5’6-FAM-ctggacaaagacactctccgcagatcat-TRMAR3’; β-actin: sense, 5’-cccagatcatgtttgagacctt -3’; antisense, 5’-gtggtggtgaagctgtagcc-3’; probe, catgtacgttgctatccaggctgtgct. Target gene expression was normalized to β-actin expression.

Costar ELISA plates (2 μg/ml) were coated with anti-human CD137 mAb (h41BB-M127, BD Pharmingen) and then blocked with 5% skim milk (BD Difco) for 2 h at 37 °C. Undiluted plasma from patients and healthy donors was incubated in the coated plates for 2 h at room temperature (RT). The captured sCD137 was detected using 2 μg/ml biotinylated-anti–CD137 mAb (4B4-1, BD Pharmingen) followed by incubation with streptavidin-HRP (1:5000, Cell Signaling Technology) for 0.5 h. The samples were developed using substrate reagents, and then stop solution (BD Biosciences) was added. The optical density (OD) was evaluated at 450 nm with a MULTISKAN GO instrument (Thermo Scientific). The plates were washed three times with 1× PBS containing 0.5% Tween-20 following every step, and all samples were assayed in duplicate. A standard curve was generated for each assay using recombinant human CD137-His (Sino Biological Inc).

Blood Tregs and CD137+ Tregs were identified with a panel of antibodies, i.e., PerCP-anti-CD3 (SK7), APC-H7-anti-CD4 (PRA-T4), BV711-anti-CD25 (2A3), BV421-anti-CD127 (HIL-7R- M21), and PE-anti-CD137 (4B4-1). To analyze the percentage of Tregs and CD137+ Tregs in tumors, single-cell suspensions were prepared by cutting fresh tumor tissues into small pieces (2-4 mm) and then using a Tumor Dissociation Kit (human, Miltenyi Biotec GmbH) and GentleMACS device according to the manufacturer’s instructions. The obtained single-cell suspensions were washed, stained for 15 min with Fixable Viability Stain 440UV, and then incubated for 30 min with a set of antibodies, including PerCP-anti-CD3, PE-Cy7-anti-CD4 (SK3), BB515-anti-CD25 (2A3), APC-anti-CD127 (HIL-7R-M21), and BV421-anti-CD137 (4B4-1). Finally, erythrocytes (RBCs) were lysed for 15 min, and the stained cells were washed twice prior to analysis with an LSRFortessa flow cytometer (BD Biosciences). To isolate CD137+ Tregs and CD137- Tregs from tumor samples, single-cell suspensions were stained with Fixable Viability Stain 440UV for 15 min, washed twice and labeled with PerCP-anti-CD3, PE-Cy7-anti-CD4, BB515-anti-CD25, APC-anti-CD127, and BV421-anti-CD137 for 30 min as described above. After RBCs were lysed and the remaining cells were washed, the resuspended cells were sorted with a FACSAria III flow cytometer.

For analysis of Tregs and CD137+ Tregs in mice, blood cells were labeled with an anti-CD16/CD32 antibody (2.4G2, mouse Fc block), washed twice and labeled with PerCP-CY5.5-CD3 (145-2C11), FITC-CD4 (RM4-5), and BV421-CD137 (1AH2) for 30 min; then, RBCs were lysed for 15 min, and the remaining cells were fixed and labeled with APC-anti-mouse FOXP3 (clone FJK-16s, eBioscience). CD8+ T cells and CD137+CD8+ T cells were labeled with PerCP-CY5.5-CD3, FITC-CD8 (53-6.7), and BV421-CD137. Mouse tumor single-cell suspensions were prepared using a mouse tumor dissociation kit (Miltenyi Biotec GmbH) and Fixable Viability Stain 575V. Dissociated single-cell suspensions were labeled with anti-CD16/CD32 and the antibodies used for blood analyses. All antibodies used for flow cytometry analyses and lymphocyte sorting except anti-FOXP3 were purchased from BD Biosciences.

For RNA-seq library preparation and sequencing, total RNA was extracted from CD137+ Tregs and CD137- Tregs using an RNeasy Micro Kit (QIAGEN GmbH). RNA quantity and purity were assessed using an Agilent 2100 bioanalyzer (Agilent Technologies). The total RNA was sent to the Beijing Genomics Institute for library construction following standard protocols. The library products were sequenced using a BGISEQ-500 instrument (BGI-Shenzhen, China). Standard bioinformatic analysis was performed by the Beijing Genomics Institute. Briefly, gene expression levels were quantified with RSEM (v1.2.12) after read cleaning and genome mapping. Differentially expressed genes (DEGs) were identified via DEseq2. Heatmaps were drawn using pheatmap (v1.0.8) according to gene expression levels in different samples. Genes with a Q value <0.05 were considered significantly differentially expressed. For immune repertoire library preparation and sequencing, 1 µg RNA samples were partitioned to construct a library for T cell receptor β (TRB) chain sequencing through a two-step PCR method. TRB genes were sequenced on a HiSeq 4000 instrument (Illumina, La Jolla, CA) using the standard paired-end 150 and paired-end 100 protocols. Base calling was performed according to the manufacturer’s instructions. The clone type frequency was analyzed using VDJ tools.

Tumor infiltrating lymphocyte (TIL) subtypes were assessed in situ by multiplexed quantitative immunofluorescence (QIF) in TMA format using the PANO 6-Plex IHC Kit (Panovue, Beijing, China, 0003100100) according to the manufacturer’s instructions. Briefly, histological sections from the patients were deparaffinized and subjected to antigen retrieval; the sections were then stained with anti-CK antibodies to detect tumor epithelial cells; anti-CD4, anti-CD8, anti-Foxp3, and anti-CD137 antibodies to detect T lymphocytes; and 4′,6-diamidino-2-phenylindole (DAPI) to visualize the cell nuclei. The following primary antibodies were used: pan-CKs (ab27988, clone AE1/AE3, 1:500 dilution), CD4 (ab133616, clone EPR6855, 1:500), CD8 (ab17147, clone C8/144B, 1:200), Foxp3 (ab20034, clone 236A/E7, 1:200) and CD137 (ab252559, BLR051F, 1:500). The TMA slides were scanned with a Vectra multispectral microscope (Akoya Biosciences). Images of single-color-stained tissue sections were obtained, and the spectrum of each fluorophore was extracted to create the required spectrum for multiplex immunohistochemical staining, which was performed in InForm2.4.8. The densities of Tregs (FOXP3+ cells), CD137+ Tregs (CD137+FOXP3+ cells) and CD137+CD8+ T cells were calculated, and the data are presented as the percentage of positive cells among total cells in each selected TMA core. The stained TMA slides were visually examined by a pathologist, and areas with staining artifacts were excluded.

A rat anti-mouse CD137 agonist (1D8, US patent #7,754,209 B2, SEQ ID No. 101 and 103) was engineered with chimeric mouse IgG2a (mIgG2a) (Wt-mAb). mIgG2a with FcγR mutations (Mut-mAb; D265A, N297A, L234A, L235A and P329A) was generated to eliminate N-linked glycosylation; these simultaneous mutations effectively reduce Fc binding to FcγR and antibody-dependent cell-mediated cytotoxicity (23). To obtain a potential local target in tumors using CD137, we prepared CD137×hEGFR (mouse CD137 and human EGFR)-bispecific antibodies with Wt-mAb and Mut-mAb formats using a eukaryotic expression system (GenScript Co. Ltd.; Supplementary Figure 4A ). Two milligrams of each purified antibody was obtained using Protein A affinity chromatography.

Female BALB/c and C57BL/6 mice were purchased from Beijing Vital River Laboratory Animal Technology Co. Ltd. All mice were maintained under specific pathogen-free conditions, housed at a controlled temperature and humidity and under a 12 h light/dark cycle. In the tumor treatment studies, 6- to 8-week-old female mice were subcutaneously implanted with 5×10⁵ CT26 cells (colon carcinoma cells, ATCC) or 1×10⁶ Lewis lung carcinoma cells (LLC cells,ATCC). Tumor area was monitored 6-7 days after tumor transplantation, and the mice were randomly divided into three treatment groups so that the mean tumors of the groups were similar. The mice were intratumorally injected with Wt-mAb or Mut-mAb (5 μg per mouse, diluted with 1×PBS) or PBS as a control 3 times at 2-day intervals. The tumor area was then measured every 3 days using electronic calipers. Blood and tumors were harvested on day 13 to analyze T cell subsets before and after treatment. The mice were sacrificed when the tumor area reached 225 mm². The percentage of mice that survived to the end point was calculated. Each experiment was repeated three times.

Data analyses were performed using GraphPad Prism 6.0. The Mann–Whitney test was used to compare non-normally distributed data between two groups, and unpaired t test with Welch’s correction was used to compare normally distributed data with unequal variances between two groups. Unpaired t test was used to compare normally distributed data with equal variances. One-way ANOVA followed by Tukey’s post-hoc multiple comparisons test was used to compare normally distributed data with equal variances between multiple groups. The Kruskal–Wallis test followed by Dunn’s post-hoc test was used to compare normally distributed data with unequal variances. OS at the end point was compared using Kaplan–Meier estimates and the log-rank test, and P < 0.05 was considered statistically significant.

Considering that sCD137 negatively regulates CD137 signaling, we initially investigated the level of free sCD137 in the peripheral blood of 59 untreated lung cancer patients and 91 healthy donors. The sCD137 concentration ranged from 2.47 to 3431.59 pg/ml (mean=121.85 pg/ml, median=17.80 pg/ml) and 4.91 to 270.58 pg/ml (mean=16.37 pg/ml, median=9.974 pg/ml) in cancer patients and healthy donors, respectively. sCD137 levels in the peripheral blood of lung cancer patients were 7.4 times higher than those in the peripheral blood of healthy controls ( Figure 1A , P<0.0001, median ± IQR). The sCD137 level was not correlated with age, pathological type or clinical stage ( Table 1 ). Then, sCD137 levels in 24 lung cancer patients who received neoadjuvant immunochemotherapy were tested, and it was found that there was a significant difference in baseline sCD137 levels between patients with a pathological complete response and patients without a pathological complete response patients. Patients with low baseline sCD137 levels were more inclined to achieve a pathological complete response ( Figure 1B , P=0.0049, median ± IQR). Twenty-four patients had 18 squamous cell carcinoma, 5 had adenocarcinoma, and 1 had sarcomatoid carcinoma; 20 were stage III, and 4 were stage IB~IIB.

To explore the source of sCD137 in resting lymphocytes, sCD137 mRNA was cloned by RT–PCR. The primers are shown in Supplementary Figure 1 (the sequences are unlined with a wavy line). We obtained two products after PCR amplification ( Figure 2A , 24 h lane). Nucleotide sequencing revealed a larger product with a molecular weight of 408 bp and a smaller product with a molecular weight of 276 bp. The shorter region contained a 132-base deletion (414 to 545, indicated by the black brackets, Supplementary Figure 1 ), which was predicted to encode the cysteine-rich domain IV (CRDIV) and STP region of membrane-bound CD137 (mCD137) ( Figure 2B ). This deletion resulted in a frame shift, generating a “taa” translational stop codon 96 nucleotides downstream ( Supplementary Figure 1 ). This CD137 isoform consisted of 169 amino acids and excluded the transmembrane region found in sCD137 ( Figure 2B ). When the RT–PCR products were inserted into the T-easy vector for sequencing, 8 of 9 colonies were identified as mCD137, and 1 colony was identified as sCD137. The full-length sequence of sCD137 and the predicted amino acid sequence are shown in Figures 2C, D .

sCD137 expression was then analyzed using ARMS-QPCR. The primers and probe were confirmed to specifically amplify sCD137 ( Figure 2E ). The expression of sCD137 mRNA in Treg (△CT =17.02) and non-Treg (△CT =19.46) subsets was directly compared (P=0.0082; Figure 2F ). The data showed that in resting T cell subsets, Tregs expressed sCD137 ( Figure 2G ), but CD4+ T, CD8+ T, and CD4+CD25+ T cells showed no sCD137 expression.

Few blood-derived CD4+ T cells expressed CD137 in healthy controls and lung cancer patients ( Supplementary Figure 2 ). CD137 expression in the Treg population in the peripheral blood was analyzed ( Figure 3A ). FACS analysis revealed that the percentage of Tregs (CD4+CD25+CD127^low) among all CD4+ T cells ranged from 3.4% to 15.6% (mean, 8.29%) in 29 lung cancer patients and 5.2% to 12.5% (mean, 7.34%) in 34 healthy controls. The percentage of peripheral blood Tregs was not significantly higher in patients than in healthy controls (P=0.1279, Figure 3B ). However, the percentage of CD137+ Tregs ranged from 1.9% to 15.2% in patients (mean, 6.06%) and from 1.2% to 9.1% (mean, 3.62%) in healthy donors; thus, the percentage of CD137+ Tregs was significantly increased in lung cancer patients compared to healthy controls (P=0.0015, Figure 3C ). No correlation was found between the percentage of Tregs or CD137+ Tregs and the clinical stage or pathological type ( Supplementary Table 1 ).

The proportions of Tregs and CD137+ Tregs in fresh tumor tissues from 10 patients were analyzed by FACS ( Figure 4A ). The characteristics of these 10 patients and the Treg and CD137+ Treg cell proportions in tumors are shown in Table 2 . The percentages of Tregs and CD137+ Tregs ranged from 10.8% to 30.45% (mean, 19.13%) and from 6.6% to 12.3% (mean, 9.47%), respectively. Interestingly, the percentages of Tregs and CD137+ Tregs were significantly increased in tumor samples compared to blood samples (Tregs: 19.13% vs. 8.286%, P=0.0008; CD137+ Tregs: 9.47% vs. 6.05%; P=0.0049; Figures 4B, C ).

CD137+ Tregs and CD137- Tregs from 3 individual tumor tissues were sorted by FACS. Treg-related gene expression was analyzed by transcriptome sequencing. The heatmap of the PI3K-Akt signaling pathway shows highly ranked differences between CD137+ Tregs and CD137- Tregs (Q value=0.0038; Figure 4D ); furthermore, Akt expression was 2.59 times higher in CD137- Tregs than in CD137+ Tregs. The expression of essential Treg-associated markers, including transcription factors such as Foxp3, CD25, CTLA-4, TIGIT and LAG3, was further upregulated in CD137+ Tregs ( Figure 4E ). Coexpression of CD25, CTLA-4, and Foxp3, which are related to effector Tregs (eTregs), was higher in CD137+ Tregs than in CD137- Tregs ( Figure 4F ), and the expression of GITR, CD27, CD137 and OX40 was upregulated to a greater extent in CD137+ Tregs ( Figure 4G ). Furthermore, CD137+ Tregs expressed higher IL-10 and Ebi3 levels than CD137- Tregs (1.3 times and 7 times higher, respectively, Figures 4H, I ). T cell receptor (TCR) sequencing analysis of three samples revealed single-clone expansion in the CD137+ Treg population ( Figures 4K, L ). In these three samples, LY108 expression was significantly higher in CD137+ Tregs than in CD137- Tregs (1.5 times, Figure 4J ).

TMA samples from 82 qualified patients were subjected to QIF with a Vectra multispectral microscope. Of the 82 patients, 35 had adenocarcinoma, 47 had squamous cell carcinoma, 49 were in stage I+II, 33 were in stage III+IV, 65 had died by Jan 2020, and 17 were alive in Jan 2020. The patients’ clinical characteristics are shown in Supplementary Table 2 . The density and spatial distribution of the CD4+, CD8+, Foxp3+, CD137+Foxp3+, and CD137+CD8+ T cell subsets are shown in Figures 5A–E .

We calculated OS from the date of surgery to death. Analysis of the association between OS and CD137+ Treg (CD137+FOXP3+) cell density in the tumor microenvironment revealed that patients with a higher CD137+ Treg density had a worse OS (P=0.0714, Figure 5F ). Regarding tumor microenvironments with a high density of activated CD137+CD8+ T cells, patients with a high CD137+ Treg density had a worse OS than those with a low CD137+ Treg density (P=0.043, Figure 5G ). Moreover, the mean IF intensity of CD137 was significantly higher in CD137+ Tregs than in CD8+ T cells (P=0.0005, Figures 5H–J ). Similarly, patients with a high Treg (FOXP3+) density in the tumor microenvironment had a worse OS ( Supplementary Figure 3 ).

To analyze whether CD137+ Treg depletion can enhance antitumor immunity in solid tumors, chimeric rat anti-CD137×hEGFR (mouse CD137 and human EGFR)-bispecific wild-type IgG2a antibody (Wt-mAb) and Fc-silenced mutant IgG2a antibody (Mut-mAb) were generated ( Supplementary Figure 4A ). We verified the anti-CD137 Fc activity of the bispecific mAbs. The two mAbs retained equivalent specific binding to mouse CD137 ( Supplementary Figure 4B ). The therapeutic potential of these CD137 mAbs was evaluated in two mouse tumor models, an LLC cell transplantation model and a CT26 cell transplantation model. The Wt-mAb exerted a considerable therapeutic effect in the CT26 cell transplantation model, as 80% of the mice were tumor free at the long-term survival endpoint; in contrast, the Mut-mAb delayed the growth of CT26 cells but did not eradicate the tumors ( Figure 6A ). Both mAbs failed to confer a significant long-term benefit in the LLC cell transplantation model, but the tumors grew more slowly in the Wt-mAb-treated group ( Figure 6B ). The percentages of Tregs and CD137+ Tregs were significantly higher in the tumor microenvironment than in peripheral blood ( Figures 6C, F ). Importantly, the total number of Tregs was decreased in both models ( Figures 6D, G , left) and was significantly decreased in the colon carcinoma model after Wt-mAb treatment (P=0.007, Figure 6D , left); however, the CD137+ Treg/Treg ratio was unchanged ( Figures 6D, G , right). Treatment did not markedly affect the percentage of CD8+ T cells ( Figures 6E, H ).